search

broadinstitute / infercnv

Inferring CNV from Single-Cell RNA-Seq
Other
557 stars 164 forks source link

Row clustering error? #558

Open J-Nicholson1 opened 1 year ago

J-Nicholson1 commented 1 year ago

Hi - thanks for a great tool

I'm having an issue somewhere between step 14&15. There wasn't an error that I could spot and the code ran through until completion - but the output plots from 15 onwards are clearly wrong, looks to be an issue with the row clustering - any ideas?

command: infercnv_obj = CreateInfercnvObject( raw_counts_matrix=counts_matrix, annotations_file=annotations, delim="\t", gene_order_file="gencode_v19_gene_pos.txt", ref_group_names=c("Ast","ExN","oRG","RG")) infercnv_obj = infercnv::run(infercnv_obj, cutoff=0.1,
out_dir=paste0("GICs@R_inferCNV/", patients[i]), cluster_by_groups = FALSE, plot_steps=TRUE, num_threads=8, denoise=TRUE, HMM=TRUE)

infercnv 14_invert_log_FC

infercnv 15_tumor_subclusters

GeorgescuC commented 1 year ago

Hi @J-Nicholson1 ,

This occurs because the subclustering is way too fragmented. In the new version, subcluster information is also plotted to help identify these issues early, and subclusters are separated by a horizontal black bar. Because there are so many subclusters, the whole figure gets covered in those bars are makes the figure unreadable. You can check the video tutorial to see how to inspect and tune subclustering, but you will basically need to set the leiden_resolution parameter lower. The process should not take long because the results from step 14 will be reloaded and only the subclustering will need to be rerun.

Regards, Christophe.